IEEE P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s Ethernet 2nd Task Force review comments

Transcription

1 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 000 SC 0 P 0 L 0 Brown, Matt PICS in Annexes 135B to 135G and 136B are incomplete. Complete PICS. Applied Micro Cl 000 SC 0 P 0 L 0 Nowell, Mark # 93 # 67 A number of specification specification values were adopted in baselines and colored magenta to represent that they were values which should be considered as Bs but the current value used was a good starting point unless further analysis suggested changing it. If after the completion of 1.1 ask Force Review, any of these magenta values have not been commented on or modified, then suggest to convert them to black font to represent that they are no longer considered Bs. his will not limit and ability to comment and adjust these values during further reviews or ballots. hey will be dealt with consistently with all other specification values in the document. Cisco Change all magenta fonts values that have not been modified at the close of 1.1 comment review to black font. Cl 000 SC 0 P 0 L 0 Brown, Matt # 80 Most skew specifications for the various new sublayers specified in clause 131 to 140 are currently marked in magenta text. For 50G sublayers, use numbers specified for corresponding 40G sublayers in For new 100G sublayers, use numbers specified for corresponding 100G sublayers in For new 200G sublayers, use numbers specified for corresponding 200G sublayers in P802.3bs. Remove related editor's notes. Applied Micro Cl 000 SC 0 P 0 L 6 udek, Mike # 150 We have defined 100GBAS-P to represent the PMA for PAM4 100G. It would be good to change the names of the PHY's that use that PMA to P instead of R Change 100GBAS-CR2 to 100GBAS-CP2 100GBAS-KR2 to 100GBAS-KP2 100GBAS-SR2 to 100GBAS-SP2 100GBAS-R to 100GBAS-P YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 000 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC 0 SR RR: Clause, Subclause, page, line Page 1 of 43

2 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 000 SC 0 P 105 L 32 Anslow, Pete R Many of the new PICS statements do not have the appropriate entries in the Support column. If the Status is "M", then there should just be "Yes [ ]" in the Support column. If the Status is "", then there should just be "Yes [ ]" and "No [ ]" in the Support column. If the Status is conditional on something else and M, then there should just be "Yes [ ]" and "N/A [ ]" in the Support column. If the Status is conditional on something else and, then there should be "Yes [ ]", "No [ ]", and "N/A [ ]" in the Support column. Ciena Scrub the New PICS statements to apply the rules in the comment. # 6 Cl 000 SC 0 P 197 L 20 # GBAS-CR4 (and 200GBAS-KR4) use the same PM control function as the 100G and 50G PHYs, which includes PAM4 precoding request. If PAM4 is requested then the PM "shall cause the adjacent PMA tp transmit... with precoding", and similarly for receive. But the 200G PHYs use clause 120 PMA which does not include precoding capability. Precoding is useful not only for controlling error propagation but also for enabling alternative receiver architectures which may be favorable in highly dispersive channels (such as cable assembly and backplane). Implementation of PMs, especially with breakout capability (where 200G, 100G, and 50G using the same circuitry), may become more complex if the 200G PMA alone does not support precoding. Suggested change is to add optional precoding to the clause 120 PMA for usage with CR4/KR4 PMs. Bring in and amend it by changing the structure that of (itle "PAM4 encoding", subclause for Gray coding including the existing text, new subclause for precoding based on ). Support for precoding in 200GBAS-R PMA should be optional, and is required for PMAs adjacent to 200GBAS-CR4 or 200GBAS-KR4 PMs. Add new control variable definitions. Add control variable mappings in and expand MI register definitions in h through n from two to four lanes. Cl 001 SC P 36 L 39 Marris, Arthur xplain what 50GAUI and 100GAUI Cadence esign Syste In the definition section point out that 50GAUI and 100GAUI carry FC encoded data while LAUI and CAUI do not. # 2 Also point this out in Annex 135A YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 001 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 2 of 43

3 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 001 SC a P 36 L 1 udek, Mike # 151 It seems strange to insert 100GBAS-R between 100GBAS-CR10 and 100GBAS- KP4. It would make more sense to insert it between 100GBAS-LR4 and 100GBAS- SR2. Also to have 100GBAS-KR2 after 100GBAS-KR4 while 100GBAS_CR4 is between Make 100GBAS-R become a1 100GBAS-SR2 become a2 100GBAS-KR2 become a Cl 045 SC P 63 L 48 Wertheim, ded here are no MI registers to configure the training protocol presets. Add the following MI registers: Preset 3 1st pre-cursor coefficient: Reserved Reserved preset_3_cm1 = 5 (c(-1) ratio -0.25) preset_3_cm1 = 4 (c(-1) ratio -0.2) preset_3_cm1 = 3 (c(-1) ratio -0.15) preset_3_cm1 = 2 (c(-1) ratio -0.1) preset_3_cm1 = 1 (c(-1) ratio -0.05) preset_3_cm1 = 0 (c(-1) ratio 0) Mellanox echnologie # 100 Cl 001 SC P 37 L 17 udek, Mike here are two four-lane versions. # 152 Replace "a four-lane version (CAUI-4, GAUI-4)" with "two four-lane versions (CAUI-4, GAUI- 4)" Cl 031B SC 31B.3.7 P 308 L 17 Marris, Arthur Replace B on lines 17 and 25 Make these the same as 100G, that is 394 and Cadence esign Syste # 3 Preset 3 2nd pre-cursor coefficient: Reserved Reserved Reserved preset_3_cm2 = 4 (c(-2) ratio 0.1) preset_3_cm2 = 3 (c(-2) ratio 0.075) preset_3_cm2 = 2 (c(-2) ratio 0.05) preset_3_cm2 = 1 (c(-2) ratio 0.025) preset_3_cm2 = 0 (c(-2) ratio 0) Preset 3 post-cursor coefficient: Reserved Reserved preset_3_c1 = 5 (c(1) ratio -0.25) preset_3_c1 = 4 (c(1) ratio -0.2) preset_3_c1 = 3 (c(1) ratio -0.15) preset_3_c1 = 2 (c(1) ratio -0.1) preset_3_c1 = 1 (c(1) ratio -0.05) preset_3_c1 = 0 (c(1) ratio 0) YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 045 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 3 of 43

4 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 045 SC P 49 L 10 Anslow, Pete R In ables 45-9, 45-10, and I Std 802.3bq-2016 has inserted a row for 40GBAS- below the row for 40GBAS-FR. Ciena Change the editing instructions for the 50G insertions to be below 40GBAS- # 4 Cl 045 SC P 63 L 41 Marris, Arthur Cadence esign Syste Remove editors note Add 50G, 100G PAM4 to text as modified by 802.3bs # 1 Cl 045 SC d P 55 L d has been updated in P802.3bs draft. # 124 Change "he transmitter, receive direction, is the transmitter that sends data towards the PCS." Cl 073 SC 73.3 P 76 L 49 "see 73-9" should be "see 73.9". correct per comment # 29 to "he transmitter, receive direction, is the transmitter that sends data towards the MAC." Cl 045 SC e P 57 L e has been updated in P802.3bs draft. # 125 Change "he transmitter, receive direction, is the transmitter that sends data towards the PCS." to "he transmitter, receive direction, is the transmitter that sends data towards the MAC." Cl 080 SC P 85 L 4 Anslow, Pete An item should be added to for the 1 lane MI for 100GBAS-R Ciena # 12 Show item g) as changing to: "he MIs as specified in Clause 89 for 40GBAS-FR and Clause 140 for 100GBAS-R use a single lane data path." Cl 082 SC P 94 L 38 Brown, Matt ditor's note has served it's purpose. Remove editor's note. Applied Micro # 83 YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 082 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 4 of 43

5 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 082 SC P 95 L 9 Anslow, Pete In the table in (renumbered as ) the entries in the support column are incorrect in the base standard. Since this table is being changed in this draft, these should be corrected. Ciena # 5 In the row for *AN1 add "No [ ]" in underline font in the support column. In the rows for AN2 through AN4 add "N/A [ ]" in underline font in the support column. Cl 091 SC 91.5 P 99 L 1 udek, Mike R # 174 he 100G Phy's call out clause 91 FC but there is no call out in those clauses as to which FC is used. here may also be other changes needed in clause 91 for exceptions. ither Amend clause 91 to explicitly add Clauses 136, 137, 138, 140, and annexes 135F and 135G (or the PHY and AUI names) with any amendments necessary (eg in section maybe in , maybe in definitely in etc.) or. Write a FC subsection for the 100G versions to go into each of these clauses describing which FC is used and any exceptions to clause 91. Cl 091 SC P 98 L 12 # 206 Figure he new optional state "CMP_AM" does not have any functions associated with it (i.e. the box is empty). efine and add the appropriate functions into the "CMP_AM" state. he new function would be similar (but not identical) to AMP_CMPAR defined in Cl 091 SC P 98 L 15 # 207 Figure I am not sure that "amp_valid and!amp_valid" are the correct exit conditions for the "CMP_AM" state. amp_valid as defined in just checkes that the received 64-bit block is a valid alignment marker payload, whereas in this state we also need to check that it is the correct alignment marker payload for the specific FC lane being tested. Need to define a new version of "amp_valid" that is set to true if the received 64-bit block matches the expected alignment marker payload. Also in this case I think that amp_valid has to be based on more than checking 64 bits (as the first 64 bits received on each FC lane are identical? Late comment: his comment was submitted after the ask Force review closed. Cl 091 SC P 98 L 27 # 208 Figure Why are variables "first_pcsl" and "current_pcsl" being used in a FC synchronization state diagram. he definitions in do not seem to apply here. I would have thought that the FC synchronization state machine would be operating of FC lanes (as defiend in ) and not PCS lanes? Perhaps "first_pcsl" and "current_pcsl" should be changed to "first_fecl" and "current_fecl" with new definitions. he definition should be changed to a variables that hold the " FC lane number " and not the "PCS lane number" as in the current definition. Late comment: his comment was submitted after the ask Force review closed. Late comment: his comment was submitted after the ask Force review closed. YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 091 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 5 of 43

6 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 091 SC P 98 L 33 # 176 Figure Is the defintion of the "AMP_CMPAR" function correct? It is not clear if "AMP_CMPAR" only compares a single 64 bit alignment maker payload (and if so which one), or a sequence of 64 bit alignment marker payloads. Figure 91-4 indicates that the start of each FC lane is comprised of 5 x 64 bit alignment marker payloads, e.g. FC Lane 0 starts with amp_tx_0, amp_tx_4, amp_tx_8, amp_tx_12 and amp_tx_16". Which of these are used by the "AMP_CMPAR" function? he situation is further confused by the fact that "AMP_CMPAR" refers to comparing values of "PCS lanes" rather than "FC lanes", even though the output of the "AMP_CMPAR" function is essentially to drive a FC lane mapping (i.e. FC_lane_mapping<x> fec_lane) No proposed solution. iscuss in task force. Late comment: his comment was submitted after the ask Force review closed. Cl 091 SC P 98 L 39 # 175 Figure he "2_Good" state is not consistent with the original Clause 91. I think it may have been copied from Clause 119 by mistake. In Clause 119 there are no FC lanes. Change "pcs_lane_mapping<x> pcs_lane" to "FC_lane_mapping<x> fec_lane" Cl 091 SC P 98 L 45 # 177 It is not clear that there are four instances of the FC alignment state machine running (one per FC lane). Add a note to make it clear that there are four instances of the "FC synchronization state machine" shown in Figure 91-8, compared to only a single instance of the "FC alignment state machine" shown in Figure 91-9 Late comment: his comment was submitted after the ask Force review closed. Cl 091 SC P 99 L 27 # 178 Need to define more clearly what is meant by "alignment markers" in the definition for "amp_bad_count". For example there are only four "amp_bad_count" counters (one for each FC lane), but there are 20 alignment markers. oes alignment marker mean a signle 66 bit PCS lane alignment marker, a single 64 bit amp_tx_x (Figure 91-4) or a collection of five 64 bit amp_tx that occur at the start of each FC lane? No proposed solution. iscuss in task force. Late comment: his comment was submitted after the ask Force review closed. Late comment: his comment was submitted after the ask Force review closed. YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 091 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 6 of 43

7 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 131 SC P 107 L 10 Reading bullet "2c" it could be interpreted that LAUI-2 can use Annex 135/. # 179 Reword to make it clear that LAUI-2 uses Annex135B/C and 50GAUI-2 uses Annex 135 /. Something like: "he PMA service interface, which, when physically implemented as LAUI-2 at an observable interconnection port uses a 2-lane data path as specified in Annex 135B or Annex 135C and when physically implemented as 50GAUI-2 (50 Gb/s twolane Attachment Unit Interface) uses a 2-lane data path as specified in Annex 135 or Annex 135" or change the text for bullet 2c to add the words "as appropriate" at the end so "he PMA service interface, which, when physically implemented as LAUI-2 and 50GAUI-2 (50 Gb/s two-lane Attachment Unit Interface) at an observable interconnection port, uses a 2-lane data path as specified in Annex 135B, Annex 135C, Annex 135 or Annex 135, as appropriate" Late comment: his comment was submitted after the ask Force review closed. Cl 131 SC P 107 L 10 Ghiasi, Ali R # 104 LAUI-2 and 50GAUI-2 are introduced to this point the reader does not know what they till they read page 113 We either need to add explicit definition for LAUI-2 is an optional 2 lanes electrical interface above the FC operating at GBd and 50GAUI-2 is an optional 2 lanes electrical interface below the FC operating at GBd. his wording should in this section or it could added in front material. Ghiasi Quantum LLC Cl 131 SC P 109 L 13 FC is mandatory for all PHYs. # 180 Change: "An FC sublayer specified in Clause 134 is available for all 50GBAS-R PHYs" to "50GBAS-R PHYs use the FC sublayer specified in Clause 134". his makes the description consistent with Late comment: his comment was submitted after the ask Force review closed. Cl 131 SC P 109 L 19 # 181 here is no mention of FC in this section? For example "he 50GBAS-R PMA performs the mapping of transmit and receive data streams between the PCS and PMA via the PMA service interface, and the mapping and multi-plexing of transmit and receive data streams between the PMA and PM via the PM service interface" he 50GBAS-R PMA also performs the mapping of transmit and receive data streams between the FC and PMA via the PMA service interface. Change the sentence to read " he 50GBAS-R PMA performs the mapping of transmit and receive data streams between the PCS and PMA via the PMA service interface, the mapping of transmit and receive data streams between the FC and the PMA via the PMA service interface, and the mapping and multi-plexing of transmit and receive data streams between the PMA and PM via the PM service interface" Late comment: his comment was submitted after the ask Force review closed. YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 131 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 7 of 43

8 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 131 SC P 114 L 21 awe, Piers R # 74 We need to go back to the principles in to work out the Skew and Skew Variation limits. Applies to 50G and 100GBAS-P PHYs such as 100GBAS-SR2, 100GBAS-R ake into account that the unit interval here is 38 ps not 97 ps, the number of lanes is 2 not 10, some PMs are serial so can't add Skew or SV, and the Skew from a possible 2-lane 40/80 km WM PM may not be the same as for a 4-lane 80 km WM PM that P802.3ba considered. Also whether there are now cost-sensitive 50G applications for which support of 40 km, maybe even 10 km, is irrelevant. ake care to round the right thing: if the buffers have to be twice as long as the SV, and if we want them in whole UI, SV should be rounded up to the next 0.5 UI not 1 UI. Mellanox Cl 132 SC P 119 L 44 # 182 I didn't think we supported LPI for 50G PHYs ( deep sleep mode is not supported)? No proposed solution. Late comment: his comment was submitted after the ask Force review closed. Cl 134 SC P 133 L 33 Brown, Matt # 81 Given that there are both "FC lanes" and "PCS lanes", the full term should be used rather than just "lanes". Where "lanes" is referring to FC lanes, replace "lanes" with "FC lanes" as necessary. Where "lanes" is referring specifically to PCS lanes, replace "lanes" with "PCS lanes" as necessary. Some specific locations: page 133, line 33, "FC lane" page 134, lines 16 and 32, "PCS lane" page 135, Figure page 138, line 5 page 141, Figure Applied Micro Cl 134 SC P 136 L 52 Slavick, Jeff # 120 he AM marker pad bit is defined to alternate between 1 and by sets it's AM marker pad bit to always be 0. A 0 is an indicator of a "Control" block and the AM 66b blocks are constructed as Control blocks. o be consistent with 802.3by remove "or 1 in an alternating pattern" from the last paragraph on page 136. Broadcom Limited Cl 134 SC P 137 L 24 Ghiasi, Ali Ghiasi Quantum LLC R x_scrambled no clear change to Start of tx_scrambled data # 105 YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 134 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 8 of 43

9 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 134 SC P 139 L 16 Slavick, Jeff # 121 rror marking is part of the decoder. he error marking pattern in Clause 91 is not appropriate for a single 257b AM block. You need to use the text from Clause 108. Add an exception to the decoder section stating the error marking is done as follows and copy the 4th paragraph of as the new text. Broadcom Limited Cl 134 SC P 140 L 13 Brown, Matt ditor's note has served it's purpose. Remove editor's note. Applied Micro Cl 134 SC P 141 L 50 # 84 # 183 It would be better if Figure was relocated to appear before sub-section Move Figure Late comment: his comment was submitted after the ask Force review closed. Cl 134 SC P 142 L 9 Brown, Matt ditor's note has served it's purpose. Remove editor's note. Applied Micro Cl 134 SC P 142 L 44 Brown, Matt # 85 # 82 he redefinition for fec_optional_states includes the opening sentence "Boolean variable that is true if the optional states are implemented and false otherwise." For the Clause 134 FC, this sentence is out of context since the "optional states" are always implemented. elete "Boolean variable that is true if the optional states are implemented and false otherwise." Applied Micro Cl 135 SC P 150 L 11 # 184 "he PMA allows the PCS (see Clause 133 and Clause 82) to connect in a mediaindependent way with a range of physical media. " Why is there no mention of FC here? he PMA also allows the FC sub-layer (see Clause 91 and Clause 134) to connect in a media-independent way with a range of physical media. Why do we single out the PCS but not mention FC? No proposed solution. Late comment: his comment was submitted after the ask Force review closed. YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 135 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 9 of 43

10 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 135 SC P 151 L 13 # 185 Figure We should decide whether to use "FC" or "RS-FC" in these SI reference models, and then be consistent across all clauses ecide whether to use "FC" or "R-FC" for the SI reference models and be consistent across all Clauses. Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 152 L 28 # 186 Figure Suggest extending Figure to show LAUI-2 interface between 50G PCS and FC, and CAUI-n between 100G PCS and FC, to better align with the subsequent text which talkes about both LAUI-2 and CAUI-n. Add LAUI-2 and CAUI-n to Figure Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 153 L 12 CAUI-4 should be CAUI-10. # 187 Replace "CAUI-4 is specified Clause 83 and associated annexes." with "CAUI-10 is specified Clause 83 and associated annexes." Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 153 L 12 Item 3) of item g) is describing CAUI-10. Change "CAUI-4" in item 3) of item g) to "CAUI-10". Cl 135 SC P 154 L 30 Figure "z" can also be 20 for 100GBAS-P. # 127 # 188 Replace "z = 4 for 100GBAS-P" with "z = 4 or 20 for 100GBAS-P" in the diagram. If you want to be more precise you could also indetify PCSL and FCL so soemthing like "z = 4 FCLs or 20 PCSLs for 100GBAS-P" and "z = 2 FCLs or 4 PCSLs for 50GBAS-R" Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 156 L 1 # 189 Why is the text starting with "In the x direction, the PMA transfers." repeated from section (page 155, line 10). Same comment for the text starting with "In the Rx direction, if the symbol is " on line 10. Propose deleting this text and perhaps the whole of section At the very least we appear to be mixing the definition of the PMA service interface and the description of the funtions within the PMA sub-layer (which belong in 135.5). Could also delete the same text in Late comment: his comment was submitted after the ask Force review closed. YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 135 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 10 of 43

11 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 135 SC P 156 L 27 # 190 It is not clear what the word "divisors" means in the following sentence " As described in , the number of input lanes and the number of output lanes for a given PMA are divisors of 2 (below the FC) or 4 (above the FC) for 50GBAS-R, or 4 for 100GBAS-P, which are the number of PCSLs/FCLs for the respective PHYs". A retimer PMA would have the same number of input lanes as output lanes, in which case I don't see how the divisor can be 2 (or 4)? Reword to make it clear what is meant by "divisors". Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 156 L 38 # 191 he list starting on line 38 is missing the condition "Whether the PMA is adjacent to the FC" Update the list to include "Whether the PMA is adjacent to the FC" Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 157 L 37 Note LAUI-2 is missing from notes "a" and "b" in Figure # 192 Change "a If 50GAUI-n or 100GAUI-n immediately above this PMA" to "a If LAUI-2, 50GAUI-n or 100GAUI-n immediately above this PMA" and change "b If 50GAUI-n or 100GAUI-n immediately below this PMA or if this is the closest PMA to the PM" to "b If LAUI-2, 50GAUI-n or 100GAUI-n immediately below this PMA or if this is the closest PMA to the PM" Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 157 L 50 Missing reference to LAUI-2 # 193 Change "If the interface between the sublayer below the PMA and the PMA is physically instantiated as 50GAUI-n or 100GAUI-n, the PMA..." to "If the interface between the sublayer below the PMA and the PMA is physically instantiated as LAUI-2, 50GAUI-n or 100GAUI-n, the PMA..." Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 158 L 7 # 194 "he bit multiplexing behavior is illustrated in Figure " If the bit muxing behavior is a detail of the more generic PMA functional block diagram, then I suggest it would be better for Figure to come after It is a bit confusing the way it is currently structured where a diagram of some internal detail of the PMA comes before the high level PMA functional block diagram. Move Figure to after Figure (i.e. reorder the figues). Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 158 L 11 # 195 here are no PCLS below the FC (or if they are then the number is 4 and not 2) so the text is somewhat confusing. Change "he number of PCLS/FCLs z is 2 (below the FC) and 4 (above the FC) for 50GBAS-R interface and 4 for 100GBAS-P interfaces" to "he number of PCSLs/FCLs z is 2 FCLs (below the FC) and 4 PCSLs (above the FC) for 50GBAS-R interface and 4 FCLs (below the FC) for 100GBAS-P interfaces" Late comment: his comment was submitted after the ask Force review closed. YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 135 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 11 of 43

12 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 135 SC P 158 L 12 # 196 "he nominal bit rate Rlane of each PCSL/FCL is Gb/s for 50GBAS-R above the FC and." his is incorrect. he nominal bit rate for the 50GBAS-R PCS lane is Gb/s as described in Clause 133. here are also no FCLs above the FC. Change "he nominal bit rate Rlane of each PCSL/FCL is Gb/s for 50GBAS-R above the FC and " to "he nominal bit rate Rlane of each PCSL is Gb/s for 50GBAS-R above the FC and " his wording is still a bit cumbersome and could be improved further. Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 158 L 18 # 197 he following sentence is a bit cumbersome "he Baud rate is equal to half of the bit rate when the number of physical lanes is 1 for 50GBAS-R or the number of physical lanes is 1 or 2 for 100GBAS-P (PAM4 symbols are sent or received on the lanes)"> his text, or similar, seems to be repeated several times in the clause. Why not simply state that "the Baud rate is equal to half the bit rate when PAM4 encoding is implemented". It is already stated elsewhere (several times) that PAM4 encoding is used when "the number of physical lanes is 1 for 50GBAS-R or the number of physical lanes is 1 or 2 for 100GBAS-P ". oo much repetition to quote a BBC radio 4 program! Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 158 L 33 # 198 "As the PCS (see Clause 133 and Clause 82) has fully flexible receive logic, an implementation is free to perform the mapping of PCSLs/FCLs from input lanes to output lanes without constraint" It is also a requirement that the FC (Clause 91 and Clause 134) has flexible receive logic as well to make this satement true. Include a reference to FC (Clause 91 and 134). Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 159 L 9 # 199 Figure he result of the equation "x+4/m" is incorrect. he correct answer should be x+1 and not 1.Same comment for equation x+4/n on line 27. Replace "x+4/m=1" with "x+4/m=x+1" and replace "x+4/n=2" with "x+4/n=x+2" Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 159 L 13 Figure Redundant set of muxes. elete the redundant set of muxes. # 200 Late comment: his comment was submitted after the ask Force review closed. YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 135 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 12 of 43

13 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 135 SC P 159 L 41 # 201 "he Skew (relative delay) between the PCSLs/FCLs must be kept within limits so that the information on the lanes can be reassembled by the PCS" his statement also applies to the FC. Change "he Skew (relative delay) between the PCSLs/FCLs must be kept within limits so that the information on the lanes can be reassembled by the PCS" to "he Skew (relative delay) between the PCSLs/FCLs must be kept within limits so that the information on the lanes can be reassembled by the PCS and FC" Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 160 L 12 (toward the PM) is redundant as transmit direction has already been defined. Remove "(toward the PM)" # 202 Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 161 L 6 Remove the reference to PCSLs. # 203 Change "while maintaining the bit order and position of PCSLs/FCLs on lanes sent in the receive direction towards the MAC." to "while maintaining the bit order and position of FCLs on lanes sent in the receive direction towards the MAC. Cl 135 SC P 162 L 54 # 204 "precoder_up_tx_enable_i, precoder_up_rx_enable_i, precoder_down_tx_enable_i,"..in these variable names do tx and rx still represent direction of data flow with respect to the PM, so for example "precoder_up_tx_enable_0" would turn on decoding for precoded PAM4 symbols recevied on lane 0 from the FC, e.g. generating G(j) from P(j)?, or to put it another way is "precoder_up_tx" an input to the PMA and "precoder_up_rx" an output of the PMA (and from/to the FC susblayer). No proposed solution. Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 163 L 23 # 30 If the PMA is adjacent to a PM, it would be beneficial for diagnostic purposes to have local loopback implemented through the PM (the PMs have no loopback control and the PM clauses refer to the PMA loopback). Also, Note 2 in regarding network disruption should really be placed here, since this is where loopback is controlled. Add the following N after the second paragraph: N 1 he intention of providing this loopback mode is to permit diagnostic or self-test functions to test the transmit and receive data paths using actual data. If the PMA is adjacent to a PM, it is recommended that the local loopback be implemented through the PM and that the signal paths that are exercised in the loopback mode encompass as much of the PM circuitry as is practical. Add the following N after the last paragraph: N 2 Placing a network port into loopback mode can be disruptive to a network. Late comment: his comment was submitted after the ask Force review closed. YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 135 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 13 of 43

14 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 135 SC P 165 L 21 # 205 here are no detailed descriptions provided for each of the MI variables in able Please see section or 91.6 as examples. Add a description for each of the MI variables in able Cl 135 SC P 167 L 1 "PMA precode request status (1.604)" is missing in able Add rows for "PMA precode request status (1.604)" to able # 130 Late comment: his comment was submitted after the ask Force review closed. Cl 135 SC P 165 L 44 # 126 he description of PMA precoder control in able is inconsistent with Clause 45. Change " " to " " Change " " to " " Change " " to " " Change " " to " " Change " " to " " Change " " to " " Change " " to " " Change " " to " " Change "precoder_up_tx_enable_1" to "precoder_tx_up_enable_1" Change "precoder_up_tx_enable_1" to "precoder_tx_up_enable_0" Change "precoder_up_rx_enable_1" to "precoder_rx_up_enable_1" Change "precoder_up_rx_enable_1" to "precoder_rx_up_enable_0" Change "precoder_down_tx_enable_1" to "precoder_tx_down_enable_1" Change "precoder_down_tx_enable_1" to "precoder_tx_down_enable_0" Change "precoder_down_rx_enable_1" to "precoder_rx_down_enable_1" Change "precoder_down_rx_enable_1" to "precoder_rx_down_enable_0" Change "PMA precoder control" for the appropriate name of "PMA precoder control x down", "PMA precoder control Rx down", "PMA precoder control x up", or "PMA precoder control Rx up". Add rows for "PMA precoder request down (1.605)". Add rows for "PMA precoder request up (1.606)". Cl 135G SC 135G.1 P 349 L 10 Ghiasi, Ali R # 116 For this clause we are referencing CL120. broken specification. C2M simulation were based on channels with ICN of ~0.7 db where QSFP28 ICN is in excess of 4 mv. For background please see attach presentation Both BS and C task force need to develop a robust C2M specifications, this will likley involve tighting the transmiter RLM and jitter and receiver sensitivity. Ghiasi Quantum LLC Cl 136 SC P 176 L 28 Brown, Matt ditor's note has served it's purpose. Remove editor's note. Applied Micro # 86 YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 136 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 14 of 43

15 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 136 SC P 177 L 1 # 31 he three paragraphs starting here describe the expected performance of a "link" in a very detailed way using normative language. But this "link" comprises multiple components - two hosts (each containing one or more PHY chips, PCB, connectors, and spanning multiple sublayers), and medium. hese components may be supplied by multiple vendors. he standard is written with the objective that a system of compliant transmitter, compliant cable assembly, and compliant receiver, will operate at the required BR (and FLR); but it is the task force's responsibility, not any single vendor's responsibility. No single vendor can guarantee a normative requirement for link performance. here are separate specifications for the transmitter, receiver, and cable assembly, and they are coupled together to facilitate the expected overall "link" performance. hese normative requirements are sufficient, and there is no need to add a system-level normative statement that no vendor is accountable for. here should be no "shall" and no PICS item for this text. Instead, it would be a service to readers if the introduction includes the expected performance of a complete physical layer (in terms of frame loss ratio or mean time between errors) and the suggested performance of a PM and an adjacent PMA (in terms of detector/bit/symbol error ratio). his applies to other PM clauses too. A detailed replacement text will be presented. Cl 136 SC P 177 L 1 awe, Piers # 73 As far as I can see, "link BR" is not defined or even referenced. he term was used in 802.3by where there is only one lane so less ambiguity, but not normatively, nor explicitly defined. Is this is the interface BR defined in ? If so, call it interface BR. Mellanox Cl 136 SC P 179 L 12 udek, Mike he inter-sublayer interface for 100G isn't defined in Replace with 80.3 Cl 136 SC P 179 L 13 he PM service interface for 100Gb/s PHYs is defined in # 153 # 128 Change the reference to in the third paragraph of to a reference to Cl 136 SC P 180 L 23 he bit time in the footnote a) of able is wrong. # 129 Change "20ns for 50GBAS-CR, 10ns for 100GBAS-CR2, and 5ns for 200GBAS-CR4" to "20ps for 50GBAS-CR, 10ps for 100GBAS-CR, and 5ps for 200GBAS-CR4". Cl 136 SC P 180 L 34 ditor's note has served its purpose. elete the editor's note. # 32 YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 136 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 15 of 43

16 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 136 SC P 180 L 48 # 33 Cl 136 SC P 181 L 17 # 34 Skew constraints are B. Skew constraints are in magenta. Based on discussions in the task force meetings and ad hoc call, it is proposed that the PM skew constraints be independent of PM type and that 50G skew constraints enable future multi-lane 50G PMs. he suggested numbers are the same as the ones on clause 80 (100 Gb/s thernet) and clause 116 (200 Gb/s, 802.3bs 2.2). Since this is a single-lane PM, an informative note can be added for the fact that PM and medium do not add skew and have no skew variation. For Skew at SP2, change B to 43 ns. For Skew at SP3, change B to 53 ns. For Skew at SP4, change B to 134 ns. For Skew at SP5, change B to 145 ns. elete the sentences "Since the signal at the (...) represents a serial bit stream, there is no Skew Variation at this point" and instead add the following informative N before the final paragraph: N--Since the signals at the PM service interface and the MI represent a serial bit stream, there is no Skew Variation at these points. he 50GBAS-CR PMs and cable assembly do not contribute to the skew between SP2 and SP5. Based on discussions in the task force meetings and ad hoc call, it is proposed that the PM skew constraints be independent of PM type. he numbers in are consistent with clauses 116 for 200G and clause 80 for 100G. Comment can be appied to all PM clauses. Change all numbers in from magenta to black. Use same skew and skew variation numbers in other clauses. Cl 136 SC P 181 L 41 # 35 Control and status variable mapping should be updated, so that the editor's note can be removed. Update table and table according to variable defiinitions in and register mapping in clause 45. Add registers in clause 45 if necessary. Implement with editorial license. elete editor's note. Cl 136 SC P 183 L 5 Brown, Matt Applied Micro ditor's note has served it's purpose. Remove editor's note. # 87 YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 136 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 16 of 43

17 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 136 SC P 183 L 6 udek, Mike # 154 he ditor's note is helpful and would be helpful for future readers of the standard. Why do we want to remove the note prior to publication? However Clause 92 (including the MI which is specified for clause 136 by reference to Clause 92) uses the 0 to 3 nomenclature not 1 to 4. It may be better to re-label the lanes here to match what is done in Clause 92. ither Change the Note from an ditor's note to a note. or as the previous paragraph already starts with "note that" just make this sentence into the last sentence of that paragraph. r remove the +1 in Figure and the labels for SL and L (throughout the clause) and the editors note. Make the same change to Clause 137 (and the editors note on page 277 line 13 Cl 136 SC P 206 L 3 udek, Mike # 160 he reference to would require the test channel meets the requirements for clause 110 not for 136. Add "except that the cably assembly shall meet the requirements of and the cable assembly test fixture shall meet the requirements of 136B.1 Cl 136 SC P 185 L 22 ditor's note has served its purpose. elete the editor's note. # 38 Cl 136 SC P 185 L 35 # 36 "ther loopback signal paths may also be enabled independently using loopback controls within other devices or sublayers" he statement is vague and arguably incorrect. he only loopback controls specified are in the PMA (referenced below). he "other loopback paths" include remote loopback, but enabling both local loopback and remote loopback together on the same PHY may yield unexpected results. his statement is within an informative note, but does not add any valuable information. elete the quoted statement. Cl 136 SC P 185 L 37 # 37 his subclause describes the _local_ loopback function. Control of the local loopback function is specified in Change the cross reference from to Cl 136 SC P 187 L 34 Ghiasi, Ali R # 106 It would be nice to have one sentence description of Fig instead of the read whole next page You could coule right something like "he output of PRBS generator is demultiplexed 1:2 into A and B. he ouput A goes through block x3 (maybe you need better name) to generate PAM2 signal. For PAM4 signal generation the output A and B are 1st gray encoded or optionally the pre-coder is enabled. Ghiasi Quantum LLC YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 136 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 17 of 43

18 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 136 SC P 188 L 36 udek, Mike # 155 Rather than defining the pre-coding here it would be better to reference it from clause 135 Reference clause and delete equation Cl 136 SC P 190 L 34 Slavick, Jeff # 122 Forcing of c(0) to zero is not the desired affect for the "No qualization" state of the cursor. Since we do refer to it as cursor and not an attenuator. Change "by forcing its value to be zero" to "forcing its value to have no equalization effect." Broadcom Limited Cl 136 SC P 190 L 34 # 39 In "no equalization" state c(0) should not be set to zero but rather to 1. setting c(0) to zero would practically create an attenuated, inverted, or zero signal. Change FRM forcing its value to zero setting it to No equalization. he No equalization value is 1 for c(0) and 0 for c( 2), c( 1), and c(1). In the coefficient update algorithm ( ) replace line 43 "ck_ask=0" with the following lines: if k = 0 ck_ask = 1 else ck_ask = 0 end if Cl 136 SC P 190 L 36 Rechtman, Zvi R # 102 he Control and status fields of the traning frame are M endcoded. When those fields have an odd number of logical ones, the encoded M is not C balanced (+6.25%), and therefore the traning frame is not C balanced. Add a parity bit to the Status field. his bit will be used to keep an even number of logcial ones and to ensure C balance of the traning frame Add new variable "paritiy bit" to the Status field as bit 7. "parity bit" - even parity bit is used to ensure that the encoded M field is C balanced. Mellanox Cl 136 SC P 191 L 42 o address the editor's note, some additional text is suggested. # 16 Add the following paragraph at the end of this subclause: "Receiver frame lock shall be set to 0 upon entering RAINING mode and shall not be set to 1 until tf_lock is true." elete editor's note. Cl 136 SC P 191 L 43 Slavick, Jeff # 118 In forced bring-up mode using link training, if both sides are in RAINING_FAIL state, and one side is reset, it could immediately start it's max_wait_timer because it's got tf_lock and if the other side is still sending "ready to respond" the rcv_tf_lock could be true good. Add the following text to "While training_failure is RU this bit is transmitted as a 0." Broadcom Limited YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 136 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 18 of 43

19 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 136 SC P 192 L 43 Slavick, Jeff he N_Q state for c(0) is 1 not 0. # 123 Replace ck_ask = 0 with "if k = 0 ck_ask = 1 else ck_ask = 0 end if" in the coef_req = N QUALIZAIN Broadcom Limited Cl 136 SC P 193 L 26 # 40 he definition of CHCK_VLAG is strictly based on whether the specified maximum output voltage would be exceeded if the request is carried out. In practice a transmitter may be unable to carry out the request due to the combination of coefficients, even though the specified maximum voltage is not reached. he transmitter may even be designed so that it is limited to a voltage strictly below the specified maximum, so that it would not exceed that maximum even if the request were carried out. he definition should be changed so that it does not address the max voltage specification but rather the transmitter equalization capability. he required capability is specified in the electrical characteristics subclauses; it does not need to be listed here. Change the function name to CHCK_Q in the definition and in the algorithm listing above. Change he definition text FRM "against the maximum steady-state voltage (see ). Returns true if the steadystate voltage would exceed the maximum. " "against the transmitter s steady-state voltage (see ) and equalization capability. Returns true if the resulting combination of coefficients would exceed the maximum steadystate voltage or the transmitter s equalization capability." Cl 136 SC P 193 L 46 # 17 he number 2 is in magenta, a peculiar color. Nothing seems wrong with this value. Paint it black. Cl 136 SC P 196 L 20 # 18 here has been no discussion based on the editor's note. he function definition does not imply immediate execution, similar to other funcitons, e.g. UPA_Cn and UPA_IC which may take some time to execute. delete editor's note. Cl 136 SC P 198 L 18 Slavick, Jeff rcv_tf_lock is used in Figure but never defined # 119 Add definition for rcv_tf_lock into Variable derived from the Reciever Frame Lock bit of the status field of the received training frames. Broadcom Limited Change the status string from "maximum voltage" to "equalization limit" in and in table YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 136 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 19 of 43

20 I P802.3cd 50 Gb/s, 100 Gb/s, 200 Gb/s thernet 2nd ask Force review comments Cl 136 SC P 198 L 35 Wertheim, ded R # 101 Reset the PM control state machine upon timeout a-synchronically with the peer state machine can create a race where each state machine assumes the peer is locked, tries to lock and fails. Add a transition from RAINING_FAIL to INIIALIZ on break_training_timer_done. Add a break_link_timer variable to imer for the amount of time to wait in RAINING_FAIL to assure that the link partner also entered a the RAINING_FAIL state. he timer shall expire 60 ms to 75 ms after being started. Set local_rx_ready <= false in the RAINING_FAIL state. Mellanox echnologie Cl 136 SC P 198 L 37 # 19 Behavior in RAINING_FAIL state is not specified to be different from other states. If training frames are still transmitted with frame lock indication, the partner may time out and reach RAINING_FAIL too; this could become a deadlock unless both sides are reset within a short period of each other. Cl 136 SC P L udek, Mike # 161 otal symbol error ratio (used in table ) is confusing. It means here FC symbol error ratio not PAM symbol error ratio Change the parameter name to "FC symbol error ratio" here and in section Cl 136 SC P 201 L 21 # 145 able has reference to specifies that differential and commonmode signal levels are measured with a PRBS9 test pattern. Change the first sentence of as follows: he transmitter on each lane shall meet the specifications given in able and detailed in the referenced subclauses excepting that the differential and common-mode signal levels are measured with a PRBS13Q test pattern (see ). his deadlock can be avoided by having the "training" variable set to false in RAINING_FAIL state, and making this value resets the training frame lock state diagram: - he "failed" device would go out of lock and signal no frame lock until it is reset by mangement; by that time, the partner will also fail. - Resetting one device would make it go to either AN signaling or, if AN is bypaeed, to SN_F, but it will not proceed to train_local because the other device does not signal tf_lock. - Resetting the second device would make both devices go to either AN or SN_F, and then they can acheive training frame lock and advance to RAIN_LCAL In figure 136-7, add inside RAINING_FAIL: "training <= False" In figure 136-8, change the open condition "reset" to "reset +!training". Cl 136 SC P 201 L 26 ditor's note has served its purpose. delete editor's note. # 20 YP: R/technical required R/editorial required GR/general required /technical /editorial G/general Cl 136 CMMN SAUS: /dispatched A/accepted R/rejected RSPNS SAUS: /open W/written C/closed Z/withdrawn SC SR RR: Clause, Subclause, page, line Page 20 of 43